Machine Learning Tutorial - Making a recommendation engine IN THE BROWSER

In part 2 of our machine learning crash course, we'll make a recommendation engine in the browser.
Here's the code!
codepen.io/wil...
We will make a recommendation engine example that fully runs in the browser with really good performance. Artificial intelligence and machine learning are awesome new frontiers in computing and data science. Web developers can use machine learning and artificial intelligence to solve problems in fun and unique ways.
Once we build a trained model using our neural network, we can quickly run thousands of possible options through our model to get ideal recommendations for our users.
~-~~-~~~-~~-~
Also watch: "Tailwind CSS - why CSS utility classes save so much time"
• Tailwind CSS - why CSS...
~-~~-~~~-~~-~

[@nus]

This is absolutely awesome, the earliest one give me the courage to dive into ML, but I immediately thought how can I do complex things, and this one filled my mind with various scenarios that I can use, thank you a lot, I had a strong desire to get into ML for 2 years but never happened until I saw your first video, this is priceless.

[@learncodeacademy]

Awesome to hear! Yes, ML is an incredible tool for your mind as well as anything else - it gives you a totally different possible approach to solving some problems...and makes you think about whole new problems to solve.

[@vsansanv29101_sd]

@@learncodeacademy facebook id can i have thanks

[@BrettCooper4702]

What I liked about this is that there wasn't a need for huge amount of predefined training data as it learned from the the user doing the star stuff. Very useful routine and I can already think of a way to apply this to one of my projects. Thanks

[@learncodeacademy]

Awesome!

[@AbhishekKumar-mq1tt]

Thank u for this awesome video, please make more video series or project on AI and machine learning with JavaScript

[@djsamke384]

This is honestly the best tut ive ever seen!!! Great work

[@balazsorban]

Amazing! Please continue with this series! It helps me to understand machine learning, finally!

[@cakeside]

Another great one. Haven't had the time yet to try it out myself but I'm getting some fun ideas from this. Thanks!

[@kevinzhang8974]

thanks for the part2 of this. I have succuessfully duplicate the result ... now, I need to dig into the code to learn the magic.

[@learncodeacademy]

I'll also be covering more detail of how neural networks actually work in future videos

[@florianbouron8340]

This series rock! Please show us more about this ML course

[@zh0r1k]

Awesome Tutorial, you make everything look simple, just like it should be. Great to see JS finally go ML with f.e. deeplearn.js (GPU rendering). Would love to see more ML videos, so keep up the great work!

[@learncodeacademy]

Yeah deeplearn is really great...not nearly as simple as brain for newbies, but really great.

[@zh0r1k]

Definitely, to understand the basics, you should start with brain.js.

[@NoMatterCode]

Really good! So.. What about self machine learning? The snake game is a good example, or chess maybe? You can probably tell to robo-musician how does it's melody sounds like. There's a couple of things to do with this great stuff.

[@itzspida111]

Can't wait for part 3!

[@montanacaleb]

Cool idea. It would be cooler to have a server building the best themes from everyone's ratings.

[@learncodeacademy]

That's the next video! :)

[@ultravidz]

Coolest series ever

[@learncodeacademy]

thanks!

[@VinsysVlogs]

Much appreciated! Waiting for the next part...

[@fahrvergnuugen]

Is it possible to use text as input dimensions? Suppose you had paragraphs of text which described an object, and you wanted to use ML to classify the object based on that text. How would you go about pre-processing the text for use in a NN?

[@MaurizioCalarese]

Keep going! Very interesting serie. Many thanks & compliments.

[@TheTiivik]

Will, thanks for showing another cool use case. I was wondering if there is a case where you don't need to normalize the datasets? Like when you had input if height + weight and output of child/adult in the first episode that were values outside of 0:1 space?

[@learncodeacademy]

In the case of weight/height, you would just normalize on expected values...so a human weighs between 0 and 500lbs, so you could divide weight by 500. and divide height in by 90 (since the tallest human is around 90in) In general, you want all of your dimensions to be on the same scale. Normalization helps to ensure performance when the neural network is using "gradient descent" to train. If you don't normalize data, you have to use a lower learning rate (a configuration option when you do new brain.NeuralNetwork(opts) ), which slows down the training. If you imagine a flat graph, and plot hundreds of points on it with x ranges from 0 - 1000 and y ranges from 0 - 1, then you get a very long, very short rectangle of possible data points. If you were to try to look at that rectangle and determine what x,y coordinates most values "clustered" at, it would be difficult. If it were a square, because x only went from 0-1, then it would be much easier to see your data and see where in that square most of the data set was landing. I hope that helps and makes sense!

[@TheTiivik]

Thanks! I think it does make sense :) So correct me if I'm wrong, in some cases training results would get even better if data was not just normalized, but also scaled so the clusters would be more clearly "visible"?

[@learncodeacademy]

Yes, scaling is really the part of normalizing that matters...you don't want dimensions to have wildly different scales. So even though brain requires 0-1, you don't want some dimensions to have a 0-.01 scale. Honestly, I probably could have even normalized my data better in that the background will be from 0-50 and the foreground will be from 50-255. my bg values could have been 0 = 0, 1 = 50, and the others 0 = 50, 1 = 255. Because the background color values will never get above .196 the way I have it now.

[@TheTiivik]

LearnCode.academy aha! Makes sense, thanks for your time explaining!

[@DevMoSofi]

You are the best. Make more videos like this, please.

[@DaveSohan]

This was very useful. Thank you so much!!

[@ThiagoVieira91]

Seems very cool. I haven't watched the video yet, but I'm feeling that will be pretty hard to follow the instructions, since I can't read the code. The font is too small.

[@normaltrader7136]

thank you for your videos, your videos have helped me a lot ! huge love from India ! keep youtubing peace

[@michaelz329]

amazing. will you also show tensorflow.js or deeplearn.js in the future?

[@learncodeacademy]

Yeah, we probably won't stick to brain for too long. I LOVE the simplicity of it (you can USE a NeuralNetwork before you fully UNDERSTAND how to build one), but those libraries are more powerful.

[@cforcoder]

Sir please make more video in this topic "MACHINE LEARNING " with neural network ... in JAVASCRIPT

Very inspiring 👍

[@sam.kendrick]

Tangential, but I love your voice! Thanks for the effort and resources!

[@learncodeacademy]

Hah, thanks! I've put some time into tweaking my mic/compressor setup over the last year.

[@advrover]

Thanks for the great video.. Please increase font size of your editor from next video.

[@DerekAGilbert]

This is awesome content. Keep it up!

[@hntrr]

Are u gonna make more vids(PLS SAY YES)

[@anthoniezz]

Hey! Could you teach Python\Java language? I have gone through all of the web development videos and almost finishing with the JavaScript videos and I would really like to learn more. Some of the channels there that "teach" Java\Python are explaining very bad, And I would like it if you could teach it.

[@kotopult]

As explanative and interesting as always :)

[@fenhirbr]

Very good! Keep going!

[@nathanbrunetti9852]

Does this need to run on a local server to work properly? I'm trying to replicate it locally and can't get it to work.

[@muhammadusman7217]

Thanks for the video 👍🏻👍🏻👍🏻👍🏻🥇

[@dudeking1000]

Is it possible to modify this code to make a game recommends engine?

[@danielvsetecka2233]

Would you review the TensorFlow.js library?

[@learncodeacademy]

We'll probably get to deeplearn.js or tensorflow.js in future videos, yes.

[@mikhailsobol3895]

Speaking of themes, which one are you using?

[@learncodeacademy]

I'm using Glance Contrast (rainglow) in Visual Studio Code

[@mikhailsobol3895]

Thanks!

[@x__dos]

why not to combine R, G, B into single dimension?

[@learncodeacademy]

because it was important that the neural network was able distinguish & add weight to individual values...for instance a bg with more green would get less weight on the g dimension for me, because I'll always downvote a green background.

[@preetsinghchhabra6117]

Amazing!! 👍🏻

[@kevinzhang8974]

how did you enter the star symbol ">★

[@learncodeacademy]

I went here www.i2symbol.com/symbols/stars to copy/paste it.

[@kevinzhang8974]

thanks for the tip!!! I collected this site to my list ...

[@desi-musk]

Genius!

[@sushiljolly4408]

Hey i m new here...can anyone tell me that i want to learn programming...so from which vedio i should start...because there are lots of vedios here

[@hakooplayplay3212]

amaizing!

[@BiranchiNarayanNayak]

The screen resolution is very poor.

[@anuraghazra4772]

const msg = "Thank You Very Much.....";

[@attention72]

Wow

[@teresasheppard9174]

attention72 [e-mail]

[@pawarvikram03]

nic #nikvikcreations

[@adamkenton2585]

would be great if you could start your projects from scratch and not have all the files prepared off camera

[@learncodeacademy]

I actually did several video renditions of this one from scratch and each time it was just confusing the main point of machine learning...it ended up almost being an html+css video. I finally opted to start from a finished state and precap what was going on. I guess in this circumstance, though, I could have coded the JS from scratch.